Abstract To overcome a challenge in the field of imaging, Content-based Image Retrieval (CBIR) is used to find digital images in large datasets. When distinct functionalities are employed separately, the majority of present imaging systems provide less accuracy. Shape, texture, and colour are examples of low-level characteristics that are used to store various sets of models in the database. Based on the query images, related categories of images are then fetched. This paper proposes the hybrid approach of different shape, texture (cartoon feature) and colour feature. Further fuse features will be selected by neighbourhood Component Analysis (NCA) for machine learning i.e. SVM training. Validation of simulation results is achieved by using several databases. Experiments have shown that the accuracy of a NCA selected features in Corel dataset is up to 96%. The simulation results show strong performance based on recall, precision, accuracy, and F-score.
Alan : Mühendislik
Dergi Türü : Uluslararası
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|